Bootstrapping Data Science at uSwitch

Presentation by Paul Lam Data Scientist @Forwardtek presentation at Data Science London 21/03/12

Transcript

1. %RRWVWUDSSLQJ 'DWD6FLHQFHDWX6ZLWFK 3DXO/DP 'DWD6FLHQWLVWX6ZLWFKFRP 'DWD6FLHQFH/RQGRQ0DUFK
2. :KDWWRGRZLWKDOOWKHVH'DWD"
3. Ɣ 6HFRQGELJJHVWSULFHFRPSDULVRQVLWHLQ8. Ɣ *%RIGDWDSHUPRQWK X6ZLWFKFRP
4.  6R:KDW'RHVD'DWD6FLHQWLVWGR"  6KRUWYHUVLRQ ,GRSURJUDPPLQJDQGVWDWLVWLFV   /RQJYHUVLRQ 7HOOPHZKDW\RXGRZLWK\RXUGDWDILUVW   
5. )LUVWIHZZHHNV RQPHHWLQJV  $VNLQJORWVRI TXHVWLRQV  8QGHUVWDQGLQJ KRZGDWDLVEHLQJXVHGRUQRWXVHG  6WDUWLQWHUURJDWLQJWKHGDWDEDVHV 
6. ,QIRUPDWLRQ)XQQHOV 'DWD %,'HYHORSHU $QDO\VW 3URGXFW0DQDJHU ,QIRUPDWLRQ 'HFLVLRQ
7. 6FLHQWLILF$SSURDFKWR6WXG\'DWD Ɣ 'DWDQRWDVDE\SURGXFWRIEXVLQHVV  Ɣ ([DPSOH%LRWHFKUHVHDUFK D &(2ZDQWVFXUHIRUFDQFHU E 5HVHDFKHUVUXQH[SHULPHQWV F 5HILQHPRGHOEDVHGRQUHVXOWVUHSHDW E XQWLO; G 'HOLYHUUHVXOWV Ɣ &(2SRLQWVGLUHFWLRQEXWGRHVQRWGLFWDWH ZKDWUHVHDUFKHUVGRZLWKWKHGDWD
8. 'DWD6LORV
9. 4XHVWLRQ +RZZRXOG\RXPHUJH\RXUGDWDLQWRXVDEOH PHVKHV LQDQDJLOHRUJDQL]DWLRQZLWK GHOLEHUDWHGDWDVLORV"   'RQ W  5HSODFHZLWKDPRQROLWKLFGDWDZDUHKRXVH  3XVKGDWDLQWR+')6ZKLOHNHHSLQJDOO GHSDUWPHQWVDXWRQRPRXVZLWKWKHLUGDWD VLORVIRURSHUDWLRQ  )XOWRQ,%093$QMXO%KDPEKUL0XVW%LJ'DWD$OWHUWKH(QWHUSULVH"5HDG:ULWH:HE0DUFK
10. 6HFRQG0RQWK&DSWXULQJGDWD
11. &XUDWLQJGDWD (7/
12. $SDFKH+LYH64/RQ+DGRRS
13. 3URJUDPPLQJDGKRFTXHULHV Ɣ &RPSRVLELOLW\ Ɣ ([WHQVLELOLW\ Ɣ 0DLQWDLQLELOLW\ Ɣ ,QWHUDFWLYLW\
14. ([DPSOH  +RZDUHZHGRLQJRQFURVVVHOOLQJ"
15. +HUH VDJUDSK &XVWRPHUIORZEHWZHHQFKDQQHOV KWWSFLUFRVFD UDQGRPGDWD 
16. &DSWXUH!&XUDWH!0RGHO!&RPPXQLFDWH 0RVWRIWKHZRUN 0RVWRIWKHUHVXOW
17. 6RPHGDWDSUREOHPVFRQVLGHUHG Ɣ 6WUXFWXULQJ ż SDUVLQJHYHQWVLQWRRUGHUHGYHFWRUV Ɣ &OHDQLQJ ż XSWRRIRXUZHEORJUHFRUGVDUHIURPERWVDQG FUDZOHUV Ɣ 9DOLGDWLQJ ż ZKHQMRLQLQJWLPHEDVHGGDWDGRQRWDVVXPHWKDW WKHWLPHVWDPSVEHWZHHQVHUYHUVDUHV\QFKURQLVHG Ɣ 0LVVLQJGDWD ż RIUHFRUGVKDYHQRXVHU,'
18. PRQWKVODWHU :KDWKDYHZHHQDEOHG"  
19. (YLGHQFHEDVHG'HFLVLRQPDNLQJ Ɣ UHJUHVVLRQDQDO\VLV ż ZKDWLVUHWXUQRQLQYHVWPHQWRI;"  Ɣ H[SHFWHGYDOXHFDOFXODWLRQ ż ZKDWLVWKHYDOXHRIGRLQJ;"  Ɣ K\SRWKHVLVWHVWLQJ ż GLGGRLQJ;PDNHDGLIIHUHQFH"  Ɣ RSWLPLVDWLRQ ż KRZWRGHOLYHUDEHWWHUH[SHULHQFHWRFXVWRPHU;"
20. ,QIRUPDWLRQLV(DV\    6ZHHW,GRQ WKDYHWRGRDQ\WKLQJ +HPDOX6ZLWFK'HYHORSHU
21. 6XPPDU\ 2XUGDWDVFLHQFHSURFHVV   +\SRWKHVL]H  &DSWXUH  &XUDWH  0RGHO  &RPPXQLFDWH